From 76fc935c34eae157c3fb967eb756abd3a7557073 Mon Sep 17 00:00:00 2001 From: jane abdo Date: Wed, 19 Jun 2024 19:00:12 -0400 Subject: [PATCH] Update README.md --- README.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/README.md b/README.md index 5fca114..41a8e78 100644 --- a/README.md +++ b/README.md @@ -8,7 +8,7 @@ Hey! I'm Jane. I have a background in neuroscience and I'm now completing my professional master's in biomedical engineering at Polytechnique Montréal. In an effort to blend both, here's my project on computational neuroscience:)

Controlling machines with imagination

- +

Introduction:

A variety of movement types can be decoded from brain signals during movement execution, ex: wrist flexion and extension, grabbing, finger moving… ( Volkova et al, 2019). These decoded signals can then be used to control external devices, such as a screen cursor, a mouse or a prosthetic limb. Certain handicapped populations, like paralyzed and amputated people, could largely benefit from the control of external devices. As they do not have brain signals associated with the execution of movement, other ways of controlling the external device are needed. Fortunately, studies have shown that motor imagery (imagining executing movement) and motor control (executing movement) share neural mechanisms, by activating similar brain regions ( Guillot et al, 2009).
Hence, the question is: Can we decode movement types based on brain signals from imagined movement? @@ -34,7 +34,7 @@ Jupyter notebook containing data processing, classifiers and data visualization

Tools :

Methods & Results:

The starting point was examining the data. Upon doing so, I realized the electrode placement was different depending on the subject. Since the study was realized in the context of an presurgerical epileptic monitoring, the location of the electrodes depended on the approximate source of the epilepsy. The 3D brains with the electrodes of each individual were plotted. Only the electrodes present in the precentral and postcentral gyruses were selected, as these regions are involved in the execution and imagination of movement (INSERT REF). - + Click here for the interactive version: After that, the performance of the classifier on each individual was plotted. @@ -46,9 +46,9 @@ The following steps were about getting a better classification. Three classifier The three classifiers were compared on both the actual and the imagined movement conditions.
For the actual movement: - +
For the imagined moveemnt: -

Conclusion :


Contrary to my beliefs, it seemed that no single classifier excels for everyone; each individual's data requires a tailored approach for optimal performance.
Fortunately, each individual had at least one classifier with an accuracy > 50%.